Microsoft Technology Licensing, LLC

Congyong Su

Li Huang

Christopher Lee Mullins

Houdong Hu

A computer-implemented technique uses a machine-trained model to convert input information provided by a user into a natural language expression and a semantic vector. The semantic vector provides a distributed representation of the natural language expression. The technique then uses the generated semantic vector to find a target data record having a target semantic vector that is within a prescribed distance to the generated semantic vector. The technique then outputs at least a target natural language expression that is associated with the target data record. The target data record is produced in an offline process that converts one or more attribute values in an existing structured data record into the target semantic vector and the target natural language expression. Also described herein is a technique for training the machine-trained model and for updating existing collections of data records using the machine-trained model.

Creating and interacting with data records having semantic vectors and natural language expressions produced by a machine-trained model

Examples of the present disclosure describe systems and methods for returning resources from an isolated collection in response to a query. In some instances, the results of the query may return a large number of resources. As such, the returned results may be paginated so that a select number of results are returned. When a subsequent query is executed on the isolated collection, one or more cursors or skip tokens are used to ensure a subsequent pagination of results may be returned.

Multi-level data pagination

A technique is described herein for transforming audio content into images. The technique may include: receiving the audio content from a source; converting the audio content into a temporal stream of audio features; and converting the stream of audio features into one or more images using one or more machine-trained models. The technique generates the image(s) based on recognition of: semantic information that conveys one or more semantic topics associated with the audio content; and sentiment information that conveys one or more sentiments associated with the audio content. The technique then generates an output presentation that includes the image(s), which it provides to one or more display devices for display thereat. The output presentation serves as a summary of salient semantic and sentiment-related characteristics of the audio content.

Transforming audio content into images

In the present disclosure, systems, methods and devices for dynamically extending a federated graph are provided. Data may be received in a variety of unrelated forms and one or more classification types may be associated with received data. Relationships that link new data input to existing data sets and their associated classification types may be generated and new data input may be incorporated and stored with existing data sets according to generated data relationships. A data set may be sorted, and resources associated with a data set may be retrieved, based on dynamic query types.

Growing community of inventors

Sammamish, WA, United States of America

Congyong Su